build_index.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from sklearn.cluster import KMeans"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(202599, 128)\n"
     ]
    }
   ],
   "source": [
    "embeddings = np.load('embeddings.npy')\n",
    "print(embeddings.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/anmihajlov/anaconda3/lib/python3.6/site-packages/sklearn/metrics/pairwise.py:257: RuntimeWarning: invalid value encountered in sqrt\n",
      "  return distances if squared else np.sqrt(distances, out=distances)\n",
      "/home/anmihajlov/anaconda3/lib/python3.6/site-packages/sklearn/metrics/pairwise.py:257: RuntimeWarning: invalid value encountered in sqrt\n",
      "  return distances if squared else np.sqrt(distances, out=distances)\n",
      "/home/anmihajlov/anaconda3/lib/python3.6/site-packages/sklearn/metrics/pairwise.py:257: RuntimeWarning: invalid value encountered in sqrt\n",
      "  return distances if squared else np.sqrt(distances, out=distances)\n",
      "/home/anmihajlov/anaconda3/lib/python3.6/site-packages/sklearn/metrics/pairwise.py:257: RuntimeWarning: invalid value encountered in sqrt\n",
      "  return distances if squared else np.sqrt(distances, out=distances)\n",
      "/home/anmihajlov/anaconda3/lib/python3.6/site-packages/sklearn/metrics/pairwise.py:257: RuntimeWarning: invalid value encountered in sqrt\n",
      "  return distances if squared else np.sqrt(distances, out=distances)\n",
      "/home/anmihajlov/anaconda3/lib/python3.6/site-packages/sklearn/metrics/pairwise.py:257: RuntimeWarning: invalid value encountered in sqrt\n",
      "  return distances if squared else np.sqrt(distances, out=distances)\n",
      "/home/anmihajlov/anaconda3/lib/python3.6/site-packages/sklearn/metrics/pairwise.py:257: RuntimeWarning: invalid value encountered in sqrt\n",
      "  return distances if squared else np.sqrt(distances, out=distances)\n",
      "/home/anmihajlov/anaconda3/lib/python3.6/site-packages/sklearn/metrics/pairwise.py:257: RuntimeWarning: invalid value encountered in sqrt\n",
      "  return distances if squared else np.sqrt(distances, out=distances)\n",
      "/home/anmihajlov/anaconda3/lib/python3.6/site-packages/sklearn/metrics/pairwise.py:257: RuntimeWarning: invalid value encountered in sqrt\n",
      "  return distances if squared else np.sqrt(distances, out=distances)\n",
      "/home/anmihajlov/anaconda3/lib/python3.6/site-packages/sklearn/metrics/pairwise.py:257: RuntimeWarning: invalid value encountered in sqrt\n",
      "  return distances if squared else np.sqrt(distances, out=distances)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "KMeans(algorithm='auto', copy_x=True, init='k-means++', max_iter=1200,\n",
       "    n_clusters=450, n_init=10, n_jobs=-1, precompute_distances='auto',\n",
       "    random_state=None, tol=0.0001, verbose=0)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "K = 450\n",
    "clf = KMeans(n_clusters=K, max_iter=1200, n_init=10, n_jobs=-1)\n",
    "clf.fit(embeddings)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('data/index/centroids', 'w') as fout:\n",
    "    for i in range(K):\n",
    "        cur_centroid = clf.cluster_centers_[i, :]\n",
    "        print(' '.join(cur_centroid.astype('str')), file=fout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "N, D = embeddings.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "from collections import defaultdict\n",
    "\n",
    "def inverse_index(labels):\n",
    "    result = defaultdict(list)\n",
    "    for i, label in enumerate(labels):\n",
    "        result[label].append(i)\n",
    "        \n",
    "    return result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "centroid2idx = inverse_index(clf.labels_)\n",
    "\n",
    "with open('data/index/centroid2idx', 'w') as fout:\n",
    "    for i in range(K):\n",
    "        print(' '.join(map(str, centroid2idx[i])), file=fout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "filenames = sorted(list(os.listdir(path='data/img_align_celeba/')))\n",
    "\n",
    "with open('data/index/filenames', 'w') as fout:\n",
    "    print(' '.join(map(str, filenames)), file=fout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "Q = 512 # descriptor size\n",
    "random_directions = np.random.randn(D, Q)\n",
    "norm = (random_directions ** 2).sum(axis=0)\n",
    "random_directions /= (norm ** 0.5)\n",
    "biases = np.random.uniform(low=-0.05 / 3, high=0.05 / 3, size=(Q,))\n",
    "\n",
    "with open('data/index/directions', 'w') as fout:\n",
    "    for i in range(Q):\n",
    "        print(' '.join(map(str, random_directions[:, i])), file=fout)\n",
    "        \n",
    "with open('data/index/biases', 'w') as fout:\n",
    "    print(' '.join(map(str, biases)), file=fout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "shifted_embeddings = embeddings.copy()\n",
    "for i in range(K):\n",
    "    shifted_embeddings[centroid2idx[i], :] -= clf.cluster_centers_[i]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(202599, 512)\n"
     ]
    }
   ],
   "source": [
    "descriptors = (np.dot(shifted_embeddings, random_directions) + biases) > 0\n",
    "print(descriptors.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def bool2int(x):\n",
    "    y = 0\n",
    "    for i, j in enumerate(x):\n",
    "        if j:\n",
    "            y += int(j) << i\n",
    "            \n",
    "    return y\n",
    "\n",
    "def to_binary(values):\n",
    "    D = len(values)\n",
    "    result = []\n",
    "    for i in range(0, D, 64):\n",
    "        result.append(bool2int(values[i:i + 64]))\n",
    "    \n",
    "    return result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(\"data/index/descriptors\", 'w') as fout:\n",
    "    for i in range(N):\n",
    "        print(' '.join(map(str, to_binary(descriptors[i]))), file=fout)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}